Instances›AWS EC2›d3›d3.8xlarge
Compute
vCPUs32
Memory256 GiB
Mem / vCPU8.00 GiB
ProcessorXeon Platinum 8259
Clock speed3.1 GHz
Architecturex86_64
Storage
Type24 × 1,980 GB NVMe SSD
Bandwidth5,000 Mbps
Throughput625 MB/s
IOPS30,000
Networking
Performance25 Gigabit
Max ENIs3
IPv6Yes
GPUNone
Family sizes
d3.xlarge4 vCPU · 32 GiB
d3.2xlarge8 vCPU · 64 GiB
d3.4xlarge16 vCPU · 128 GiB
d3.8xlarge32 vCPU · 256 GiB
d3.8xlarge

The d3.8xlarge is a storage-optimized AWS EC2 instance: 32 vCPU and 256 GiB of memory, 8.00 GiB per vCPU. In the name, "d" is dense local storage, "3" is the generation and "8xlarge" is the size within the d3 family. It runs on Intel Xeon Platinum 8259 (Cascade Lake) at 3.1 GHz, and AWS introduced it in December 2020. d3 is the current generation of the d family, so there is no newer AWS size to migrate to. No single Spot node covers 32 vCPU and 256 GiB, so the equivalent is a fleet: 2 × mh.vs1.2xlarge, 32 vCPU and 256 GiB in total, $4.00/hr on AWS in US East (N. Virginia), from $0.042/hr on Rackspace Spot. Spot delivers compute as managed Kubernetes, so a node pool of that size is the native shape of the answer rather than a workaround — provided the workload scales horizontally.

Match qualityScale-out match

d3.8xlarge gives you 32 vCPU and 256 GiB in a single machine. The largest comparable Spot node, mh.vs1.2xlarge, is 16 vCPU and 128 GiB, so the equivalent here is 2 nodes rather than one.

Spot runs as managed Kubernetes, so a 2-node pool is the normal way to run things rather than a workaround. If your workload already sits behind a load balancer, a queue or a scheduler, it spreads across them without changes. If it needs all 32 cores and 256 GiB in one process, it does not.

No local diskd3.8xlarge comes with 24 × 1,980 GB NVMe SSD attached to the machine. Spot nodes use network storage only, so that disk doesn't come with it. If you were using it as fast scratch space, you'd move to a persistent volume or object storage, which are slower to read.
Pricing by region
1 hr
RegionAWS on-demandAWS Spot1-yr Reserved3-yr ReservedRackspace SpotYou save
US East, N. Virginia$3.995$1.113$2.517$1.726$0.042
$3.954 savings
US Central, Chicagono AWS region$0.042
US Central, Dallasno AWS region$0.262
Hong Kong$0.042
Sydney, Australia$5.008$0.953$3.155$2.163$0.042
$4.966 savings
London, UK$5.118$0.614$3.224$2.211$0.262
$4.856 savings
US West, San Jose$0.004
US Central, Dallas 2no AWS region$0.484
When to use it
Good for
  • ●Streaming ingest that spools to local disk
  • ●Elasticsearch and OpenSearch data nodes
  • ●NoSQL stores such as Cassandra, ScyllaDB and MongoDB
  • ●Workloads that fit inside 256 GiB of RAM
  • ●Jobs that want 24 × 1,980 GB of local NVMe scratch space
  • ●Network-heavy services — AWS rates this size at 25 Gigabit
  • ●Anything that shards cleanly across 2 nodes, which is how it maps onto Rackspace Spot
Avoid if
  • ●The working set needs more than 256 GiB of RAM
  • ●More than 8.00 GiB of RAM per vCPU is needed — that ratio is fixed for the whole family
  • ●The work is single-threaded. Most of these 32 vCPU would sit idle.
  • ●GPU or FPGA acceleration is needed
  • ●Workloads that never touch local disk. The instance-store premium is wasted.
$2,917/mo on AWS on-demand against $31/mo on Rackspace Spot for the 2-node equivalent — market-priced, not list-priced.
  • Spot sells spare capacity at market prices. It is the cheapest compute on the internet.
  • Prices are set by supply and demand. You pay what the market decides.
  • Compute is delivered as fully managed Kubernetes. All infrastructure is taken care of.
  • Egress costs $0.01/GB after the free allowance. AWS charges $0.09/GB.
Build your Cloudspace
Frequently asked questions